Sentiment Classification in Under-Resourced Languages Using Graph-Based Semi-Supervised Learning Methods

Yong REN; Nobuhiro KAJI; Naoki YOSHINAGA; Masaru KITSUREGAWA

doi:10.1587/transinf.E97.D.790

Open Access
Sentiment Classification in Under-Resourced Languages Using Graph-Based Semi-Supervised Learning Methods

Yong REN, Nobuhiro KAJI, Naoki YOSHINAGA, Masaru KITSUREGAWA

Full Text Views

102

Cite this

Free PDF (1.4MB)

Summary :

In sentiment classification, conventional supervised approaches heavily rely on a large amount of linguistic resources, which are costly to obtain for under-resourced languages. To overcome this scarce resource problem, there exist several methods that exploit graph-based semi-supervised learning (SSL). However, fundamental issues such as controlling label propagation, choosing the initial seeds, selecting edges have barely been studied. Our evaluation on three real datasets demonstrates that manipulating the label propagating behavior and choosing labeled seeds appropriately play a critical role in adopting graph-based SSL approaches for this task.

Publication: IEICE TRANSACTIONS on Information Vol.E97-D No.4 pp.790-797

Publication Date: 2014/04/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1587/transinf.E97.D.790

Type of Manuscript: Special Section PAPER (Special Section on Data Engineering and Information Management)

Category

Authors

Yong REN
  The University of Tokyo
Nobuhiro KAJI
  The University of Tokyo
Naoki YOSHINAGA
  The University of Tokyo
Masaru KITSUREGAWA
  National Institute of Informatics

Keyword

sentiment classification, graph-based semi-supervised learning

Cite this

Copy

Yong REN, Nobuhiro KAJI, Naoki YOSHINAGA, Masaru KITSUREGAWA, "Sentiment Classification in Under-Resourced Languages Using Graph-Based Semi-Supervised Learning Methods" in IEICE TRANSACTIONS on Information, vol. E97-D, no. 4, pp. 790-797, April 2014, doi: 10.1587/transinf.E97.D.790.
Abstract: In sentiment classification, conventional supervised approaches heavily rely on a large amount of linguistic resources, which are costly to obtain for under-resourced languages. To overcome this scarce resource problem, there exist several methods that exploit graph-based semi-supervised learning (SSL). However, fundamental issues such as controlling label propagation, choosing the initial seeds, selecting edges have barely been studied. Our evaluation on three real datasets demonstrates that manipulating the label propagating behavior and choosing labeled seeds appropriately play a critical role in adopting graph-based SSL approaches for this task.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.E97.D.790/_p

Copy

@ARTICLE{e97-d_4_790,
author={Yong REN, Nobuhiro KAJI, Naoki YOSHINAGA, Masaru KITSUREGAWA, },
journal={IEICE TRANSACTIONS on Information},
title={Sentiment Classification in Under-Resourced Languages Using Graph-Based Semi-Supervised Learning Methods},
year={2014},
volume={E97-D},
number={4},
pages={790-797},
abstract={In sentiment classification, conventional supervised approaches heavily rely on a large amount of linguistic resources, which are costly to obtain for under-resourced languages. To overcome this scarce resource problem, there exist several methods that exploit graph-based semi-supervised learning (SSL). However, fundamental issues such as controlling label propagation, choosing the initial seeds, selecting edges have barely been studied. Our evaluation on three real datasets demonstrates that manipulating the label propagating behavior and choosing labeled seeds appropriately play a critical role in adopting graph-based SSL approaches for this task.},
keywords={},
doi={10.1587/transinf.E97.D.790},
ISSN={1745-1361},
month={April},}

Copy

TY - JOUR
TI - Sentiment Classification in Under-Resourced Languages Using Graph-Based Semi-Supervised Learning Methods
T2 - IEICE TRANSACTIONS on Information
SP - 790
EP - 797
AU - Yong REN
AU - Nobuhiro KAJI
AU - Naoki YOSHINAGA
AU - Masaru KITSUREGAWA
PY - 2014
DO - 10.1587/transinf.E97.D.790
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E97-D
IS - 4
JA - IEICE TRANSACTIONS on Information
Y1 - April 2014
AB - In sentiment classification, conventional supervised approaches heavily rely on a large amount of linguistic resources, which are costly to obtain for under-resourced languages. To overcome this scarce resource problem, there exist several methods that exploit graph-based semi-supervised learning (SSL). However, fundamental issues such as controlling label propagation, choosing the initial seeds, selecting edges have barely been studied. Our evaluation on three real datasets demonstrates that manipulating the label propagating behavior and choosing labeled seeds appropriately play a critical role in adopting graph-based SSL approaches for this task.
ER -